The associations between insulin, hypoglycemia, and dementia: Combating threats to internal validity in a series of population-based cohort studies

Abstract

Although the association between type 2 diabetes and dementia is recognized, findings from the epidemiology literature on the effect of insulin and one of its side effects, hypoglycemia, are less clear. The currently available observational studies assessing these associations suffer from a wide range of methodological limitations that diminish their internal validity and lead to contradictory evidence. The aim of this thesis is to implement design and analysis techniques to combat bias and confounding in previous studies and to further extend knowledge on the risk of dementia associated with four interconnected diabetes-related exposures, each assessed in a separate study: 1) severe hypoglycemia, 2) age of severe hypoglycemia, 3) insulin use, and 4) the mediating effect of severe hypoglycemia from insulin use. Herein, a series of cohort studies were conducted using population-based health administrative data (1996-2018) from British Columbia, Canada housed by Population Data BC. First, we identified individuals newly diagnosed with type 2 diabetes between 01 January 1998 and 31 December 2016. Each cohort was then designed based on the research question, wherein exposure was defined accordingly. For studies 1 and 2, the exposure of interest was severe hypoglycemia compared to no hypoglycemia. For studies 3 and 4, the exposure of interest was insulin initiation compared to initiating a non-insulin class. For all cohorts, the outcome of interest was all-cause dementia. Confounding adjustment techniques including inverse probability of treatment weighting (IPTW) were used in all studies. In each study, a wide range of sensitivity analyses were conducted to ensure the robustness of results. Findings from study 1 confirm the previously reported higher risk of all-cause dementia with severe hypoglycemia after implementing exposure density sampling, a lag period, and IPTW (HR 1.83; 95% CI 1.31-2.57). Findings from study 2 show that the increased risk of dementia observed in study 1 is consistent whether hypoglycemia occurs in midlife (HR 2.85; 95% CI 1.72-4.72) or late life (HR 2.38; 95% CI 1.83-3.11). Conversely, findings from study 3 negate existing evidence and do not show an increased risk of dementia associated with insulin use (HR 1.14; 95% CI 0.81-1.60). Lastly, findings from study 4 indicate a potential role of severe hypoglycemia as a mediator of the association between insulin and dementia (Natural Indirect Effect HR 1.04; 95% CI 1.01-1.08). Collectively these studies provide further insight on the complex associations between insulin, hypoglycemia, and the risk of all-cause dementia to inform both clinicians and patients with type 2 diabetes on the need to prevent hypoglycemia. Importantly, these studies showcase the need for robust methodology when conducting observational studies for type 2 diabetes-related exposures.